{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:ZUKJVARIGN6IN5GY3J2LZ2WNYB","short_pith_number":"pith:ZUKJVARI","schema_version":"1.0","canonical_sha256":"cd149a8228337c86f4d8da74bceacdc07f1fea77e2dde8ffb2468030a2b5532b","source":{"kind":"arxiv","id":"2306.17806","version":1},"attestation_state":"computed","paper":{"title":"Stay on topic with Classifier-Free Guidance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.CL","authors_text":"Alexander Spangher, Elad Levi, Guillaume Sanchez, Honglu Fan, Pawan Sasanka Ammanamanchi, Stella Biderman","submitted_at":"2023-06-30T17:07:02Z","abstract_excerpt":"Classifier-Free Guidance (CFG) has recently emerged in text-to-image generation as a lightweight technique to encourage prompt-adherence in generations. In this work, we demonstrate that CFG can be used broadly as an inference-time technique in pure language modeling. We show that CFG (1) improves the performance of Pythia, GPT-2 and LLaMA-family models across an array of tasks: Q\\&A, reasoning, code generation, and machine translation, achieving SOTA on LAMBADA with LLaMA-7B over PaLM-540B; (2) brings improvements equivalent to a model with twice the parameter-count; (3) can stack alongside o"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2306.17806","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-06-30T17:07:02Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"0f26cf6d4a6ca04ec2a63060eb81378ab871b0585218f31ed28575b9bfc276ce","abstract_canon_sha256":"0d0ce4a0d86d84832435c1dcae1dc75da912b5454c9093b978bf9ecee7af49bd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:26:31.961640Z","signature_b64":"UwdT6t142q4zeL9YY2NJbEAaO3oJMh+/AjQge0N9Ld0j9TP8yDionL5Fzkvvfx25B0uTvnPwxSZ7kMo6AxGbAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cd149a8228337c86f4d8da74bceacdc07f1fea77e2dde8ffb2468030a2b5532b","last_reissued_at":"2026-07-05T06:26:31.961012Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:26:31.961012Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stay on topic with Classifier-Free Guidance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.CL","authors_text":"Alexander Spangher, Elad Levi, Guillaume Sanchez, Honglu Fan, Pawan Sasanka Ammanamanchi, Stella Biderman","submitted_at":"2023-06-30T17:07:02Z","abstract_excerpt":"Classifier-Free Guidance (CFG) has recently emerged in text-to-image generation as a lightweight technique to encourage prompt-adherence in generations. In this work, we demonstrate that CFG can be used broadly as an inference-time technique in pure language modeling. We show that CFG (1) improves the performance of Pythia, GPT-2 and LLaMA-family models across an array of tasks: Q\\&A, reasoning, code generation, and machine translation, achieving SOTA on LAMBADA with LLaMA-7B over PaLM-540B; (2) brings improvements equivalent to a model with twice the parameter-count; (3) can stack alongside o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.17806","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2306.17806/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2306.17806","created_at":"2026-07-05T06:26:31.961093+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.17806v1","created_at":"2026-07-05T06:26:31.961093+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.17806","created_at":"2026-07-05T06:26:31.961093+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZUKJVARIGN6I","created_at":"2026-07-05T06:26:31.961093+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZUKJVARIGN6IN5GY","created_at":"2026-07-05T06:26:31.961093+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZUKJVARI","created_at":"2026-07-05T06:26:31.961093+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09926","citing_title":"Sample Where You Struggle: Sharpening Base Model Reasoning via Entropy-Guided Power Sampling","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08221","citing_title":"De novo molecular generation with optical property preconditioning at the token level","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31247","citing_title":"FlexiSLM: A Dynamic and Controllable Frame Rate Spoken Language Model","ref_index":115,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29198","citing_title":"Guidance Contrastive Token Credit Assignment for Discrete Policy Optimization","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16739","citing_title":"EmoMind: Decoding Affective Captions from Human Brain fMRI","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12809","citing_title":"Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces","ref_index":96,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13595","citing_title":"Inducing Artificial Uncertainty in Language Models","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17884","citing_title":"SPREG: Structured Plan Repair with Entropy-Guided Test-Time Intervention for Large Language Model Reasoning","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB","json":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB.json","graph_json":"https://pith.science/api/pith-number/ZUKJVARIGN6IN5GY3J2LZ2WNYB/graph.json","events_json":"https://pith.science/api/pith-number/ZUKJVARIGN6IN5GY3J2LZ2WNYB/events.json","paper":"https://pith.science/paper/ZUKJVARI"},"agent_actions":{"view_html":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB","download_json":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB.json","view_paper":"https://pith.science/paper/ZUKJVARI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.17806&json=true","fetch_graph":"https://pith.science/api/pith-number/ZUKJVARIGN6IN5GY3J2LZ2WNYB/graph.json","fetch_events":"https://pith.science/api/pith-number/ZUKJVARIGN6IN5GY3J2LZ2WNYB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB/action/storage_attestation","attest_author":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB/action/author_attestation","sign_citation":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB/action/citation_signature","submit_replication":"https://pith.science/pith/ZUKJVARIGN6IN5GY3J2LZ2WNYB/action/replication_record"}},"created_at":"2026-07-05T06:26:31.961093+00:00","updated_at":"2026-07-05T06:26:31.961093+00:00"}